Those who are smug about data security have this week had their assertions torn asunder once again. The Ashley Madison hack – Dr No rather liked Henry Tudor’s tweet ‘Cromwell was my Ashley Madison. He got hacked too’ – reminds us that data said to be impregnable is in fact all too pregnable, if in the circs that’s the word Dr No is looking for. High profile hack after high profile hack tells us there is no such thing as secure data, just data yet to be hacked. Those signing up blithely to care.data may want to wonder whether the day will yet come when they will sheepishly tweet ‘care.data was my Ashley Madison. I got hacked too.’
Meanwhile, another data failure, or rather a forecast from data has turned out to be flawed. Putting aside the warmest global July on record – Dr No is still struggling with what ‘warmest global July on record’ really means – as ‘proof’ that man-made global warming is for real, we now have a report showing that the forecast rise in dementia prevalence has bombed. Like the article-of-faith rise in temperature caused by man, dementia prevalence has for decades been forecast to rise – only it now turns out it hasn’t. The report, published in the Lancet, found that in two studies, in the UK and Spain, there were significant falls in prevalence (prevalence is the ‘living with’ number), while two other studies, one in Sweden and one in Holland, found non-significant falls in incidence (incidence is the number of new cases over a time frame number). The rise in dementia prevalence forecast, championed by charities and chomped up by governments, just hit the fan.
A few days ago, a Radio 4 Weather Johnnie, on being ribbed about inaccurate forecasts, replied testily ‘the clue is in the name’. Earlier this month, Quentin Letts wondered quizzically, in a Radio 4 programme of the same name, What the Point of the Met Office? More pragmatically, what’s the point of the Met Office’s £97m new computer which – at least to this observer – appears to increase forecasters’ confidence, but not accuracy? Despite breezy assurances from Met Office spokespersons, even short term forecasts remain hit and miss, while for anything more than a few days ahead, well, all bets are off. Indeed, that’s why we don’t get BBQ summer forecasts anymore: it’s the Met Office tacitly admitting they haven’t got a clue.
The problem at the heart of forecasts based on models, especially computer models, be they for the number of people with dementia, the economy, the weather, or indeed – a Dr No favourite – how alcohol minimum unit pricing will affect drinking habits, is that they assume that not only will the future be same as the past, but also that you have captured the relevant pattern in a precise equation. This is a hazardous notion at the best of times. Even the noble laws of classical physics have been known to wobble at times: Newton’s apple goes pear-shaped in the sub-atomic world. In realms where the vagaries of human behaviour or the beat of a butterfly’s wings in Brazil have play, already pear-shaped forecasts can quickly turn into slippery bananas, putting the skids under the best laid plans.
Despite the clue-is-in-the-name nature of forecasts, Dr No is not forecasting nihilist, nor, for that matter, a dementia denier – it is still a horrible disease and a huge burden that deserves more attention. And forecasts are worthwhile, so long as we don’t get seduced by dazzling computers telling us they are ninety nine point nine percent confident, and so ourselves become over-confident. We need to remember the clue is in the name, and that things may turn out differently. Just because something is forecast to happen doesn’t mean it will happen.
Oh, and where have all the dements gone? Nowhere, because they never existed in the first place. They weren’t real, they were just a forecast.